EN
INTRODUCTION: The low-frequency part of extracellular potential, called the Local Field Potential (LFP), is a useful measure of neural systems activity. However, a direct interpretation of LFP is problematic as it is not a local measure – each electrode may record activity observed millimeters away from source. Estimation of current source density (CSD), the volume density of net transmembrane currents, has become a convenient way to deal with this problem. AIM(S): The aim of the study is to investigate the properties of kCSD method to develop a procedure which will facilitate optimal usage of the presented method in complicated experimental scenarios, for complex measurement setups etc. METHOD(S): In the study we use kCSD method which estimates the sources in a family of allowed CSD distributions of dimensionality larger than the number of measurements. To identify the parameters of the method leading to optimal source estimation, a statistical technique of cross-validation is used. We perform this study using Python programming language with several types of known (model) reference data and different electrodes setups. We employ singular value decomposition (SVD) method to study the internal properties of kCSD reconstruction. RESULTS: To examine the influence of the measurement setup on the reconstruction capability of the kCSD method we performed simulated study. We present error maps of CSD estimation which give us valuable insight into kCSD reconstruction quality. CONCLUSIONS: The quality of CSD estimation significantly depends on the measurement setup. This study enables the researchers to check how much they can trust the obtained kCSD reconstruction for a given setup and specific collection of recordings. FINANCIAL SUPPORT: This work was supported by EC-FP7-PEOPLE sponsored NAMASEN Marie-Curie ITN grant 264872, Polish Ministry for Science and Higher Education grant 2948/7.PR/2013/2, Narodowe Centrum Nauki grants 2013/08/W/NZ4/00691 and 2015/17/B/ST7/04123.